🧠 Machine Learning

Research on probabilistic graphical models, neural networks, natural language processing, adversarial ML, and classification methods.

Machine learning algorithms use experience to improve their performance on a given task. In practice, "experience" is understood as data, and performance is measured according to the task at hand. Our research in machine learning spans both fundamental methods and their applications to complex real-world problems.

Probabilistic Graphical Models (PGMs)

Bayesian Networks & Markov Networks
Research on learning the structure and parameters of Bayesian networks and Markov networks from data. Applications to density estimation, anomaly detection, classification, and generation of synthetic data.
Restricted Boltzmann Machines (RBMs)
Investigation of structural restricted Boltzmann machines for image denoising and classification. Research on the synergies between energy-based models and probabilistic graphical models.
Vine Copulas
Research on vine copula models as flexible multivariate dependency structures. Development of algorithms for learning the graph structure of regular vine-copulas from dependence data. Applications to machine learning tasks requiring accurate joint distribution modeling.
ACM TELO 2024 β†’
Copula-based Classification
Use of copula functions for classification and regression tasks. Vine copulas provide flexible alternatives to standard Gaussian assumptions for modeling complex multivariate distributions.

Neural Networks & Deep Learning

Neural Architecture Search (NAS)
Research on evolutionary and learning-based methods for automatically designing neural network architectures. Includes factorized model representations, heterogeneous multi-network architectures, and the interplay between NAS components and variation operators.
IEEE TNNLS 2023 β†’ IJCNN 2024 β†’
Generative Models (GANs & VAEs)
Research on generative adversarial networks and variational autoencoders, including their training dynamics, transferability, and optimization using gradient-based methods.
Physics-Informed Neural Networks (PINNs)
Development and analysis of physics-informed neural networks for solving partial differential equations arising in fluid dynamics. Research on hyperparameter influence, generalization behavior outside the training domain, and second-order optimization with domain decomposition.
Neural Computing 2024 β†’ PINN Balls arXiv 2025 β†’
Semi-Supervised Learning
Neuroevolutionary algorithms guided by neuron coverage metrics for semi-supervised classification. Combining evolutionary search with coverage-based objectives to improve learning with limited labeled data.
GECCO 2023 β†’
Neural Networks for Physical Systems
Application of neural networks (including NAS) to identify phase transitions in physical systems such as the Ising model. Research on the capabilities of neural networks for computational physics tasks.
MedPRAI 2025 β†’

Natural Language Processing (NLP)

Multi-label Hierarchical Classification
Development of multi-dimensional hierarchical classification methods for NLP tasks. Research on characterization, solving strategies, and performance measures for complex hierarchical label spaces.
Neurocomputing 2023 β†’
Large Language Models (LLMs) for Feature Selection
Exploration of LLMs as tools for feature selection and anomaly classification. Stacking predictions from multiple LLM models to improve accuracy and explainability in industrial anomaly detection tasks.
i-COSTE 2024 β†’ CALM 2024 β†’
ENIA Chair in AI & Language Technology
Participation in the ENIA (National Artificial Intelligence Strategy) Chair at UPV/EHU (HiTZ Center) focused on artificial intelligence and language technology. Research, teaching, and knowledge transfer in AI for language.
SEPLN-CEDI 2024 β†’

Adversarial Machine Learning & Explainability

Adversarial Attacks in Explainable ML
Comprehensive survey and research on adversarial threats against explainable machine learning systems. Identifying vulnerabilities in model explanations and developing defenses against attacks targeting interpretability.
WIREs DMKD 2025 β†’
Adversarial Class Probability Distributions
Extension of adversarial attack methods to produce adversarial class probability distributions, going beyond misclassification to manipulate the confidence scores output by classifiers.
JMLR 2023 β†’
Adversarial Examples in Audio
Research on generating and defending against adversarial examples in the audio domain. Investigation of the transferability, robustness, and perceptibility of adversarial audio perturbations.
Probabilistic Self-Explainable Neural Networks
Development of uncertainty-aware explanations through probabilistic self-explainable neural networks, providing calibrated confidence estimates for model predictions and their explanations.
arXiv 2024 β†’

Classification, Regression & Data Augmentation

Classification from Time Series
Classification of household devices from electricity usage time series; fault classification in building thermal systems (fan coil units). Research on the impact of imputation methods and temporal features.
SNAMS 2023 β†’ ICSCGE 2024 β†’
Feature Selection Methods
Filter method-based feature selection for multi-target regression problems, including scenarios with unattributed identities. Feature selection and causality explanation for anomaly classification in building systems.
Expert Systems 2024 β†’
Random Vector Functional Link Networks
Research on random vector functional link forests and extreme learning forests applied to UAV automatic target recognition. Evaluation of ensemble methods for classification under challenging conditions.
Eng. Applications AI 2023 β†’

Selected Publications